# A gap left by a fan tool's shutdown was taken over by a 7-year-old Japanese site: Funny-AI and "好き顔9選" · Web

> Source: https://stringers.io/en/web/web_funny-ai.com.html (Stringers / 什么值得做, generated 2026-11-08)
> Stringers answers one question every day: what small product can one or two people build in a few months and actually make money from. Numbers come from 15 databases we collect daily, each with source and date.
> For agents: these verdicts and the underlying databases (verified revenue, App Store / Google Play charts, ad libraries, GitHub, Product Hunt…) are queryable live over MCP — tools today / hot / products / directions / direction / data. Setup: https://github.com/komako-workshop/stringers (the user logs in at stringers.io → "配置 agent" to get a personal URL). Site index: https://stringers.io/llms.txt

> A fan-made Johnny's lookalike tool went viral, then got shut down over copyright worries 5 days later. Funny-AI caught the traffic instantly with its old site: photos, #FunnyAI and return links, spilling over via Threads into Taiwan and Korea.

[Web](https://stringers.io/en/web/index.md) · 11-6 written

Funny-AI(funny-ai.com) · funny-ai.com

# A gap left by a fan tool's shutdown was taken over by a 7-year-old Japanese site: Funny-AI and "好き顔9選"

**What it is** A long-running Japanese AI face-rating and lookalike site. In September it added a grid-generating page that pairwise-compares celebrity photos and picks your favorite 9 faces.

**Why it caught on** A fan-made version of the same tool went viral, then shut down 5 days later over copyright concerns. Funny-AI immediately absorbed the traffic through its established site. Results include photos, #FunnyAI tags and return links, and it spread to Taiwan and Korea via Threads.

**C** **How you could follow** If you already run a site with young users, add a ranking page for local fan circles in Korean, Traditional Chinese and other local markets, and launch before a given circle goes viral. Starting a new site from scratch to compete in Japan isn't worth it.

C grade: the mechanic can be copied in 2–3 hours, Japan already has a dozen clones, and the playbook has been written up publicly. Funny-AI wins on old-domain traffic and search authority that followers can't get, and trends rotate by circle every week or two, so what's left is small windows in foreign-language markets. Circulating in small circles

Global rank 245k · 90 days ago 778k

In September a Japanese fan trend took off: "好き顔9選", picking your favorite 9 faces from a pool of performers and sharing them in a 3×3 grid. The most popular fan-made tool ran for only 5 days before shutting down, and Funny-AI, an AI face-rating site run by one person in Tokyo, filled the gap with dozens of fan-circle versions. Its global rank climbed from 780k to 240k over two-plus months. The mechanic can be copied in 2–3 hours, so its edge is the established site and timing. Below: how it picked up the demand, why it was its turn, and whether you can still follow.

### Sep 23, 11:59 PM

On the evening of Sep 23, 2026, a fan named "Yukiho" (@C_jnys) posted on X: "本日9/23 23:59で公開終了します。" Her "Johnny's favorite face 9 picker maker" was shutting down that night.

She had been recommended by a friend on X only 5 days earlier, a post that got over 20k likes. In Sep 21, "ジャニーズ好き顔9選" hit Yahoo! Japan's real-time search trends. The rules were simple: 137 Johnny's performers, one screen at a time, asking "which face do you prefer?" until your TOP 9 remained. By shutdown, 72,459 responses had been submitted (figures from the creator, compiled by zawatto; 9-21 published, 10-08 updated).

She explained the shutdown clearly: no result images with photos, and no image downloads, because the photos belong to others and she couldn't take on the copyright risk. The grids people shared were screenshots they took themselves.

Two days later, on Sep 25, Threads and X began filling with a different kind of post: "私のK-POP好き顔9選 1位…9位 #FunnyAI funny-ai.com/sort/…".

### What is Funny-AI?

Funny-AI is run by Fujii Ryo, an indie developer in Shibuya, Tokyo. The site started 2019 years ago. Its original business: upload a selfie for AI diagnosis, covering face-attractiveness deviation score, animal face, which anime character you look like, and which K-POP member you look like. The site claims 34.9M diagnoses and 5.36M users in total (stats since 2022), and it appeared on a Japanese TV variety show in 2021. It has long spread these diagnoses across hundreds of landing pages: 66 K-POP groups, 38 J-POP groups, 46 anime characters. In Japan it has consistently been a top 10k site by Chrome real-user count.

The newly added `/sort` "好き顔9選メーカー" needs no AI and no photo uploads: you compare official photos in pairs and pick the TOP 9. The question bank 11-06 opens like this:

| Circle | Members (example) |
|---|---|
| K-POP men / women | Standard 185 / 167 members, expanded 323 / 434 members; also split by generation 1–5 and by HYBE / SM / JYP / YG |
| Japanese male idols | 401 members overall, a separate Johnny's page with 137 members, plus one page each for LAPONE and LDH |
| Actors, 2.5D, Takarazuka | Japanese male / female actors 144 / 155 people, 2.5D 394 people, Takarazuka 429 people, Korean actors 232 people |
| Other | Baseball players 223, comedians, voice actors, anime characters |

The Johnny's page URL is literally `johnnys-favorite-faces-top9`. Johnny's fans who lost their home when the original closed find their way here by searching.

### Why it blew up now

**The ranking inflection matches the surge.** Global site ranking (Tranco) was about 779k at the end of July, and in September it was 24 785k, barely moving. In October it was 1 668k, then 22 355k, and in November 5 245k. This ranking is a 30-day average and lags real traffic by one to two weeks, so the real rise started in late September, around when the original shut down.

**Sharing feeds back into growth.** Every share uses the same format: circle tag + `#FunnyAI` + a link back to that page. The original only let you share a string of names. Here you get a photo grid, which looks better and makes people want to post it. Every result image shared is recruiting new users.

**It takes risks the original would not.** The photos are official images, and the page only says "非公式・権利は各権利者に帰属" (unofficial; rights belong to the respective rights holders). The original author declined to show images out of copyright caution. Funny-AI accepts that risk in exchange for result images that spread better.

**Language isn't the barrier.** The page is only in Japanese, but anyone can play with photos and names. A Taiwanese user posted "9-grid of faces that boy bands like: funny-ai.com/sort/kpop/boys" on Threads, and another post "People who are interested can test here" got 1,200 views. A Korean user posted "내가 좋아하는 얼굴 테스트 남돌/여돌버전." In Chrome's real-user data, Taiwan first entered the top 10k in September, Korea first entered the top 50k, and Thailand, Indonesia, Brazil, and the UK first entered the top 500k. Japan was already in the top 10k and moved to the top 5k in September. So much of the extra September traffic came from people outside Japan.

**One circle, one wave of heat.** Other sites' data shows traffic follows whichever circle's grid is going viral. A page peaks and fades within a week or two. Funny-AI splits circles into dozens of pages, and each page can go viral again.

### Where the money comes from

I see no membership or paid tier. Shared links carry `#google_vignette`, the marker for Google AdSense interstitial ads. The page also has ads or affiliate links for books and entertainment news. The face-rating and look-alike diagnosis entries sit next to the sort pages and steer grid players back to the original business, which runs on ads. Revenue is not public, and the scale is unknown.

### Who it's competing with

Anyone can do this, and a lot of people already are:

| Competitors | Status |
|---|---|
| sukigao.net | A new site from late September with a multi-circle question bank, self-reported as 10-26 with 310k grids in total, half of them from the "anime male characters" page launched on 10-22; top 500k in Japan's Chrome user tiers |
| lovetype.org, oshi-torisetsu.com | Love compatibility test and fan-tool sites that added this as a side feature, top 50k in Japan |
| sukigao-av.com | Switched to adult themes. The author says on note they built it in 2–3 hours with AI-written code; on launch day the daily active users passed 3k 150 and it broke even that day (author's claim, 10-11) |
| EBiDAN edition, LAPONE edition, and others | Single-circle fan tools on GitHub Pages, earlier than the Johnny's version |
| picrew, sreate | General-purpose puzzle / sorter tools where users must find images and build questions themselves |

Going by Chrome user tiers, Funny-AI is the biggest of this batch: top 5 thousand in Japan, while the best of the others only reach the top 50k. The difference is not the product. It's a 7-year-old domain with a large base of older Japanese users and search weight. Searching "好き顔9選" ranks it in the top two (inferred from search results). A new site has to start from zero, and Funny-AI doesn't.

Funny-AI has not been covered on its own by the media, but the "好き顔9選" format has been written up by non-no, EMMARY and zawatto. The adult-version author publicly described the "3 days, 200k visits" approach on note. The niche is no longer a secret.

### If it were you

**C: You can learn a playbook from this, but joining this wave is not worth it.**

Verdict: the grid format has no real barrier, but Japan is already crowded. Funny-AI wins on its established traffic and search weight, which newcomers can't get. Heat runs on one- to two-week cycles per circle. Japan's wave has run from September to now, and it will get harder from here.

**Who it suits** : people who already run a site aimed at young people or fans, with some traffic and search weight (a test site, a horoscope site, a fan-tool site), targeting Korean, Traditional Chinese, Thai or Indonesian markets. Adding a sort page in the local language costs a few hours.

**Who it doesn't suit** : anyone wanting to build a new site from scratch to compete with the Japanese sites; anyone worried about photo copyright complaints (the original author stopped showing images for exactly this reason, and you would take on that risk); anyone expecting this page alone to build long-term traffic.

**How to build it** : a web page, not an app. One page per circle, with a question bank of currently trending local groups and actors. Users compare two at a time to pick a TOP9, and the result is a photo grid. The share text is fixed to "circle tag + your site-name hashtag + a link to that page." Put an entry point to your own existing features on the page to keep people around.

**How to get users** : search Threads and X for local-language words like "九宮格" or "얼굴 9" and see which circles are sharing. Publish that page in the days before it goes viral. Taiwanese and Korean users are already using the Japanese page, so demand exists. I haven't seen anyone make a big local version yet (not verified one by one).

**How long you have** : heat like this fades in a week or two per page. Japan's wave has run for over two months, from September to now. The window for spillover into other languages is probably until the end of the year.

**Signs you've misread it** : (1) Funny-AI's global ranking stops rising in mid-November and starts to fall, meaning the wave has passed its peak. (2) An independent grid site appears in Korean or Traditional Chinese and reaches the local top 50k within a month, meaning the gap is already taken. (3) Conversely, if Funny-AI gets a complaint over photo rights and removes the image feature, then "showing images" becomes a barrier nobody wants to touch, and whoever builds next will have to switch themes (anime characters, virtual streamers).

A final line for everyone building products: **In fan-made playbooks, once an insider-made tool blows up and then shuts down over copyright or lack of time, the gap goes to whoever already has traffic.** People with a site should watch closely; people without one should build a site first.

[Website ↗](https://funny-ai.com)

**Research notes: competitor table, acquisition evidence, full research report**

Competitors

| Name | Where | Launched | Scale | How it differs from this one |
|---|---|---|---|---|
| ジャニーズ好き顔9選メーカー (SUKIGAO 9) | Web | Before and after 2026-09-18, 9/23 went offline | 72,459 replies, recommended posts with 20k+ likes | Covered only one circle (Johnny's), with no photo result images. It shut down after 5 days, and Funny-AI picked up the gap it left. |
| Sukigao diagnosis sukigao.net | Web | late 2026-09 | Self-reported 310k 9 selections in total (10/26); CrUX Japan top 500k | A new site has no old-domain authority and relies on data-column SEO; Funny-AI's traffic is a full tier or more higher. |
| EBiDAN / LAPONE sukigao maker | Open source | 2026-05-27 / 2026-07 | A small circle, with some spread on TikTok | Only one agency, and most shares are plain text |
| lovetype.org, oshi-torisetsu.com | Web | 2026-09 joined 9 selections | CrUX Japan top 50k | A side feature on a test site or fan-promotion site; the question bank and traffic are both smaller than Funny-AI's |
| sukigao-av.com(R4VEN) | Web | 2026-09-25 | Self-reported DAU of 150k+ (at peak); CrUX Japan top 100k | Adult theme, built with AI programming in 2–3 hours, same playbook |
| IKACHI, FaceScore, FaceRatingAI, Fotor and other attractiveness raters | Web | Years ago | FaceScore Play has 830k installs; ikachi.org's Tranco rank rose from 790k to 550k over the same period | Only does AI scoring, which overlaps with Funny-AI's original business; no good-face sorting mechanic |
| picrew / sreate.com general sorting or puzzle tools | Web | Old site | picrew's Tranco rank is around 7,700 | Users have to build their own questions and find their own images; there is no ready-made library of idol photos |

Where users come from

- **Verified** : **Riding the hype (fan-driven clone after the original went offline)** : Johnny's version 9/19 blew up, 9/23 went offline (zawatto); Funny-AI's /sort was already in use by 9/25 at the latest (Threads, X tweet IDs converted to timestamps), and they also built the johnnys-favorite-faces-top9 page

- **Verified** **Organic sharing on X / Instagram** : share copy includes #FunnyAI and a deep link back to the site. The #funnyai hashtag on X is full of 9 posts.

- **Verified** **Threads spillover across languages (Taiwan, Korea)** : in Taiwan, @jing.update and @twinklingeating (1.2K views); in Korea, @minky0516 reposted the /sort link. CrUX shows Taiwan and Korea first entering the chart in September.

- Inferred **Search traffic from an old domain plus media explainers** : searching 「好き顔9選」 puts it #2 on the first page; non-no, EMMARY and ticketjam wrote how-to explainers.

- Inferred **Own TikTok / YouTube accounts** : there is an @funnyai_official account, but I haven't seen any hit posts.

## Funny-AI (funny-ai.com): why it took off

### One line

An old Japanese AI face-rating / look-alike diagnosis site that launched in 2019 caught the late-September 2026 wave of the Japanese fan trend "好き顔9選" (pick the 9 faces you like most from a pile of celebrity photos and assemble them into a 3×3 grid). The earliest fan-made tool that went viral, 「ジャニーズ好き顔9選メーカー」, was taken down in 9/23. Within a day or two, Funny-AI shipped a version covering dozens of fan circles, with images, a `#FunnyAI` hashtag and backlinks, and the trend spilled over to Taiwan and Korea via Threads.

### 1. What it is

- Operator: Fujii Ryo (Shibuya, Tokyo, sole proprietorship "FUNNYAI"). Site launched in 2019. Self-reported "34.9M diagnoses and 5.36M users in total" (GA4 data from 2022-04 onward). Featured on Japanese TV show "スクール革命!" 2021-10-10 (source: [funny-ai.com/about](https://funny-ai.com/about)).

- Legacy business: upload a selfie for AI diagnosis, e.g. attractiveness deviation score, handsome/ugly, cute/ugly, animal face, which anime character or Pokémon you resemble, which celebrity or K-POP member you look like. The site has a large set of programmatically built landing pages: 66 K-POP groups, 38 J-POP groups, 46 anime characters, 18 celebrity categories, plus skin diagnosis and more (source: homepage nav [funny-ai.com/](https://funny-ai.com/)).

- New format `/sort` 「好き顔9選メーカー」: no photo upload. You compare official photos pairwise and pick a TOP 9. The question bank covers K-POP boy and girl groups (standard 185/167 people, extended 323/434 people, filterable by debut generation and by HYBE/SM/JYP/YG agency), Japanese male and female idols, Sakurazaka-group, KAWAII LAB., Japanese male and female actors (144/155 people), 2.5 dimensional actors (394 people), Takarazuka (429 people), Korean actors (232 people), baseball players (223 people), comedians, voice actors and anime characters, among others (sources: [funny-ai.com/sort](https://funny-ai.com/sort), [funny-ai.com/popular](https://funny-ai.com/popular)).

- Share copy reads 「私の○○好き顔9選 1位…9位 #○○好き顔TOP9 #FunnyAI funny-ai.com/sort/...」. Searching `#FunnyAI` on X turns up many posts in this format (sources: [x.com/hashtag/funnyai](https://x.com/hashtag/funnyai), Yahoo!リアルタイム検索 「#FunnyAI」).

- Monetization: shared links carry `#google_vignette`, which means the page runs Google AdSense interstitial ads (example: [www.threads.com/@binnie_briize0904/post/DdsUNNoE9cg/](https://www.threads.com/@binnie_briize0904/post/DdsUNNoE9cg/)). The page also shows ads or affiliate links like 「解剖学本を探す」 and 「芸能ニュースを読む」. I didn't see a subscription or paywall.

- Almost all traffic is on the web. It has TikTok account @funnyai_official and YouTube account @funnyai_official for content, and I found no app.

### 2. Data (our traffic database, 2026-11-05 list)

- Tranco: 7/30 is 779,254,9/24 and 785,279 is 10/01 and 667,558,10/15 is 492,938,10/22 and 355,220,11/05 is 244,876 (mostly flat). The inflection point is early October. Tranco is a 30-day rolling average and lags real traffic by 1–2 weeks, so real traffic started rising in late September, matching the social evidence below.

- CrUX: globally, 2026-05 through 08 were all in the top 500k, rising to the top 100k in September. Japan held steady in the top 10k, rising to the top 5 thousand in September. Taiwan first entered the top 10k in September, Korea first entered the top 50k; Thailand, Vietnam, Indonesia, the Philippines, Brazil, the UK, Canada, Australia and Turkey first entered the top 500k in September.

- This suggests most of September's growth came from East Asian Chinese-speaking and Korean-speaking users outside Japan. Japan's base was already large (top 10k).

### 3. Competitors (same thing: 好き顔9選 / favorite-face rankers, plus AI face-rating diagnostics)

| Competitors | Form | Launched | Scale (our database / public data) | How Funny-AI differs |
|---|---|---|---|---|
| ジャニーズ好き顔9選メーカー (SUKIGAO 9, @C_jnys) | Fan-made web tool | Before and after 2026-09-18, 9/23 went offline | 72,459 responses before it went down; a friend's recommendation post on X got over 20k likes (quoted by zawatto) | Only one fandom (Johnny's); avoided copyright worries by sharing text only, not result images with photos; shut down within 5 days, leaving a gap |
| EBiDAN 好き顔9選メーカー (@M1kk4n71) | GitHub Pages | 2026-05-27 | Small niche | One agency only, Elo pairwise comparisons, text-only sharing |
| LAPONE 好き顔診断 | GitHub Pages | Early 2026-07 | Spread on TikTok | Only covers JO1, INI, etc.; has a 'share image to X' button |
| Sukigao diagnosis sukigao.net | Standalone site | late 2026-09 | Self-reported 311,362 total 9 選 (10/26); half came from the 「anime male characters」 page launched 10/22. CrUX top 500k in Japan; not in Tranco | Similar number of fandoms, plus a data column for SEO; the site is new and has no aged-domain authority |
| lovetype.org(Love Type) | Website | Earlier than 2026-05 | CrUX Japan rose from the top 500k to the top 50k (September) | A side 9 選 added to a love-compatibility test site |
| oshi-torisetsu.com/sukigao | Website | 2026-09 | CrUX Japan top 50k, Taiwan top 100k (newly entered in September) | Added to a fan-activity tools site |
| sukigao-av.com(R4VEN) | Adult-oriented website | 2026-09-25 | Self-reported over 3k DAU on day 150 of launch, breaking even the same day (note 2026-10-11) | Reskinned with an adult-video actress theme; self-reported AI-coded build time of 2–3 hours |
| shindan_navi, IFTI (idol-face-type9.pages.dev), ayaya0227.github.io/tobe-sukigao9, etc. | Small sites or TikTok accounts | 2026-09 | Small | Single fandom, or added a 'face type' questionnaire |
| sreate.com, dumapp.com, picrew | Generic ranking tools / collage makers | Old site | picrew's Tranco rank is around 7,700 | Users have to build their own question sets; no ready-made photo library |
| Face-rating tools: IKACHI, FaceScore (830k Play installs), FaceRatingAI, Fotor, gazoulab | Web / App | Old | ikachi.org's Tranco rank rose from 790k to 550k over the same period | Only rating; Funny-AI's legacy business competes head-on with these |

Takeaway: anyone can build the 「好き顔9選」 format, and since September there are plenty of clones. By CrUX, Funny-AI has the most traffic of the bunch: top 5 thousand in Japan, while the best of the others only reach the top 50k.

### 4. Growth tactics

1. **趁热潮接盘(已确认)** :9/19 杰尼斯版在 X 上爆了,9/21「ジャニーズ好き顔9選」进 Yahoo! 实时趋势,9/23 23:59 原作者下线(zawatto.com/meme/sukigao-9sen,2026-09-21 发、10-08 更新)。Funny-AI 的 `/sort` 最晚 9/25 已经有用户在用:Threads @annysays16 9/25 发帖;X @koroniki1209 发的 K-POP 9 選,推文 ID 换算成时间是 9/25 16:18 UTC。它还单独做了一页 `johnnys-favorite-faces-top9`,接住了原版下线后没处去的杰尼斯粉丝。

2. **Results share with branding and links (confirmed)** : share copy consistently includes `#FunnyAI` and a link to the site. Users repost it organically on X, Threads and Instagram; e.g. Instagram: 「流行ってるぽいので、私もFunny-AIの…やってみました #Funny-AI」.

3. **Cross-language spillover on Threads (confirmed)** : a Taiwanese user, "9-grid of faces that boy bands like: funny-ai.com/sort/kpop/boys" (@jing.update), and "People who are interested can test here" (@twinklingeating, 1.2K views); a Korean user @minky0516, "내가 좋아하는 얼굴 테스트 남돌/여돌버전." This matches Taiwan and Korea first entering the CrUX charts in September. The page is only in Japanese, but photos plus names are enough to play, so the language barrier is low.

4. **Old-site SEO foundation (inferred)** : the domain is 7 years old and has been in the top 10k in Japan on CrUX. Searching 「好き顔9選」 ranks it #2 on the first page (right after sukigao.net in Jina search results). Media explainers (non-no, EMMARY, ticketjam Cal-cha, iステップ) raised search volume, and the aged domain was well placed to capture it.

5. **One page per fandom, constantly adding new ones (confirmed)** : sukigao.net's data shows traffic depends on which page's result images are going viral, e.g. comedians in early October, then anime male characters from 10/22. Funny-AI split the format into dozens of pages by fandom, agency and debut generation, and each new fandom page can trigger another wave.

6. Paid acquisition: no evidence found (the domain isn't in our ad library, C2 pool), so I judge growth to be mainly organic.

### 5. Why this one (and not another clone)

- **Traffic base and trust** : it is already one of Japan's major AI face-rating diagnosis sites. Knowledge Ichiba answers include people saying 「funny-ai's attractiveness deviation diagnosis is the most accurate」, and articles like vanceai have also recommended it. A new site has to build traffic from zero; this one didn't.

- **Fast launch, deep question bank** : in the gap left by the original's shutdown, it rolled out a dozen-plus categories at once (K-POP, J-POP, actors, voice actors, sports, Takarazuka and more), groups of 100 to over 400 people, plus extended versions, so one site could catch fans from many fandoms.

- **Willing to take on photo-rights risk** : the original creator explicitly declined to provide result images with photos. Funny-AI uses official photos and includes only a one-line disclaimer 「非公式・権利は各権利者に帰属」. The results look better and are easier to share as images. Photos are hosted on img.kaoshindan.com (likely another domain run by the operator, unverified).

- **Sharing designed to bring traffic back** : `#FunnyAI` plus a deep link means every shared result image sends new visitors back to it.

- **Redirecting new visitors to the old business** : the page still links to the looks-rating and look-alike diagnostics, sending sorter traffic to the diagnostics where people upload selfies, then monetized with AdSense interstitials.

### 6. Differentiation

The product has no technical moat. Face-ranking is an old fan-circle game, and there are over a dozen clone sites; the author could build one in 2–3 hours. Its edge is "an established site's authority + the full circle question bank + result images with photos and a brand tag", and the combination is what captured the largest share of the clone wave. The real driver is **timing plus distribution** : it launched when the original had just gone offline and demand was at its peak, and it had an older site base to immediately catch search traffic. "AI" played essentially no role in this growth; sukigao 9選 doesn't use AI at all.

### 7. Visibility

- No public revenue figures, and no standalone media coverage. Media covers the "好き顔9選" game format; Funny-AI is just one of the sites that gets mentioned, e.g. non-no and ticketjam articles focus on sites like sukigao.

- The format itself is already widely discussed (the zawatto entry, non-no, EMMARY, ticketjam). R4VEN publicly described on note an approach of "3 days, 200k visits, paid back the same day", and copycats have already flooded in. Overall read: the product circulates within a small circle, and the niche is already crowded.

### 8. What to borrow

- Watch for hand-built games in fan circles: once a "community-made tool goes viral and then goes offline" pattern appears, quickly ship a full-circle version on an existing traffic site.

- Fixed share-copy format: result, circle tag, brand hashtag, deep link.

- One page per circle, with constant new additions, catching one wave of trends after another; built as programmatic SEO pages to capture search traffic over the long term.

- A language-free visual format (photo plus name) naturally spills over into Taiwan, Korea and Southeast Asia, with Threads as the main spillover channel.

- New traffic feeds the old business, monetized with interstitial ads.

### 9. Still unclear

- Exact launch date of `/sort` (we only know that 9/25 was already in use by someone; Wayback has no record).

- How much of the growth came from sukigao 9選 versus organic growth of the old diagnostics business. The similar looks site ikachi.org was also rising in the same period, so the whole category may have heated up together.

- Whether it generates share images with photos (the data card says so, while the Instagram posts show result screenshots); and whether the photo copyright risk will lead to takedowns or complaints.

- Revenue scale, and how much traffic will remain once the craze fades. sukigao's data shows single-page interest decaying within a week or two.

This article was written by AI from public data and the data we collect daily. Numbers are sourced, and inferences are labeled.